Written by Admin Alex · Fact-Checked by M.Ali · Info Verified September 2026
We review and update this article regularly as new information becomes available.
TL;DR: Nvidia CEO Jensen Huang said at a tech event in Scotland that the company’s chip sales are on track to roughly double in 2027 compared to this year, backing up guidance he gave on the August earnings call. The forecast leans on Nvidia’s new Vera Rubin platform, hundreds of billions of dollars in supply commitments, and demand that shows no sign of slowing. Wall Street analysts have already revised earnings estimates upward 39 times in the past 90 days with zero cuts.
Huang wasn’t shy about the number. Speaking overseas rather than on a scripted earnings call, he told the room that 2027 chip sales would roughly double what Nvidia sells this year, a repeat of guidance he first floated back in August. Saying it twice, in two very different settings, tends to mean a CEO believes it.
The engine behind that math is Vera Rubin, Nvidia’s next data center platform, which the company says is generating $40 billion in revenue for every gigawatt of computing capacity deployed. Compare that to Blackwell’s $25 billion per gigawatt and the jump makes sense. Nvidia is calling Vera Rubin the fastest product ramp in company history, which is saying something for a company that’s had several.
None of this happens without buyers lined up. Nvidia has locked in $279 billion in memory supply obligations just to keep pace with orders. Financial heavyweights including Apollo, BlackRock, and Goldman Sachs are reportedly mobilizing more than $500 billion combined toward AI infrastructure buildouts. Amazon Web Services alone plans to add 2 million more GPUs by fiscal 2029. OpenAI’s own compute commitments now amount to something like 12 gigawatts, a figure that would have sounded like science fiction three years ago.
Sovereign buyers are the new whales
Enterprise and sovereign AI spending is growing at roughly 100% a year, according to the numbers Huang cited. Governments building their own AI infrastructure, rather than renting it from a handful of US hyperscalers, have become one of Nvidia’s fastest-growing customer categories. That’s a meaningful shift from a few years ago when Nvidia’s revenue was overwhelmingly tied to a short list of American cloud giants.
Nvidia is projecting fiscal 2028 revenue around $678 billion, implying growth near 70%. Those are eye-watering figures for any company, let alone one that was primarily known for gaming graphics cards a decade ago.
Wall Street has taken the hint. Consensus fiscal 2028 earnings-per-share estimates jumped from $12.67 to $15.57 in the last three months alone, with 39 upward revisions and not a single downgrade. The stock responded with a 2% pop, pushing shares to around $219, within striking distance of the 52-week high of $236.
What could actually slow this down
The honest caveat here is memory. High-bandwidth memory remains the bottleneck constraining how fast Nvidia, and everyone buying from Nvidia, can actually build. Even with $279 billion locked in, that’s a supply chain with a lot of moving parts and very few alternate suppliers. Power availability for data centers is the other quiet constraint nobody wants to say out loud at a keynote.
Bottom Line: A CEO doubling down on a wild-sounding forecast in a second venue, backed by real supply contracts and real customer commitments, is different from hype. Nvidia’s numbers are big enough that skepticism is healthy, but the demand signals here, from sovereign AI buildouts to AWS’s GPU orders, look like more than a sugar high. Whether 2027 actually delivers a doubling is a bet the market has clearly decided to make alongside Huang.
